The good S3 came up to 441GH/s avg within an hour while the questionable S3 from Florida was only hitting 100GH/s. Not one with a whole lot of patience, I decided to install the 0811.bin firmware on the questionable S3. Did. Then worried a bit as it was only doing 10GH/s avg for some time. I restarted cgminer (/etc/init.d/cgminer stop, /etc/init.d/cgminer start) and there was no apparent improvement. Then I noticed there was no frequency listed on the status page so of course into the new Advanced tab on miner config, I specified the default and saved. Up 6 minutes now and in the 400's GH/s. Time will tell.
...Now over an hour and the S3 from Florida, having the 0811.bin firmware, seems to have settled out at 435.65GH/s. What would explain this and its low fan speed is if an ASIC near the LM75A has some engines turned off - perhaps it had overheated at some point and damaged one. I have noticed that the lower center ASICs under the LM75A tend to have drier thermal paste than other ASICs. Center of the hashing board. The curve of the inside heatsink fins and the direction of spin on the fans seems the reason one board, the top chain, gets hotter and has the faster fan speed.
If an ASIC near the LM75A is damaged and turns off engines, it gives the system an inaccurate assessment of the board's temperature. ASICs toward either end working properly will be running hot.
...their logic that the lower hashing board should be on the cooler side made sense so it's been torn down again, end plates reversed, fans reversed, controller board moved to the other end, and reassembled. Now coming up.
Florida, land of the floating chads.
Yes, this is better. Before it barely got up to 435GH/s with the low hashing board on the right side when looking at the RJ45 connector end. The right side is the top chain and generally has the higher temperature and faster fan. In this case it had lower temperature and slower fan than the lower chain. So, taking the end plates and their fans, swapping ends and moving the controller board, put it back the way it was shipped, the questionable board on the cooler side.
Almost right away the average went above 441GH/s and has remained above 440GH/s for the last 3h 23m and is presently 442.04 tho the utility is only 15. No hardware errors in that time. It was started as the afternoon cooled. It remains to be seen what happens tonight as the humidity increases.